Evaluation better

aufgaben/p5/pla_applied.py

@@ -4,6 +4,7 @@ import random
 from algorithm.pla.perceptron_learning_algorithm import train, train_pocket
 from algorithm.pla.perceptron import Perceptron
 from algorithm.pla.transfer_functions import normalized_tanh
+from aufgaben.p6.evaluate import evaluate
 from korpus import create_bewegung
 
 
@@ -31,35 +32,14 @@ def apply_pla():
     train(perceptron, training_data, max_iterations, learning_rate)
     print("\nPLA: Zunächst Gewichte, dann Klassifizierung. Gehen = 1, Kniebeuge = 0")
     print(perceptron.weights)
-    cnt_true_pos = 0
-    cnt_true_neg = 0
-    cnt_false_pos = 0
-    cnt_false_neg = 0
-    unclassified = 0
-    for i in range(500):
-        jog_res = perceptron.classify([joggen.sensoren[0].werte[i], joggen.sensoren[1].werte[i], joggen.sensoren[2].werte[i]])
-        knie_res = perceptron.classify([kniebeuge.sensoren[0].werte[i], kniebeuge.sensoren[1].werte[i], kniebeuge.sensoren[2].werte[i]])
-        if math.isclose(1, jog_res, abs_tol=0.1):
-            cnt_true_pos += 1
-        elif math.isclose(0, jog_res, abs_tol=0.1):
-            cnt_false_neg += 1
-        else:
-            unclassified += 1
-        if math.isclose(0, knie_res, abs_tol=0.1):
-            cnt_true_neg += 1
-        elif math.isclose(1, knie_res, abs_tol=0.1):
-            cnt_false_pos += 1
-        else:
-            unclassified += 1
-    print(f"""
-    Anzahl Prüfungen: 1000
-    True Positive: {cnt_true_pos}
-    False Positive: {cnt_false_pos}
-    True Negative: {cnt_true_neg}
-    False Negative: {cnt_false_neg}
-    Unclassifiable: {unclassified}
-    Fehlerrate: {(cnt_false_neg+cnt_false_pos) / 10}% bzw. {(cnt_false_neg+cnt_false_pos+unclassified) / 10}%
-    """)
+    amount_of_weights = 3
+    correct_data = []
+    incorrect_data = []
+    for i in range(amount_of_weights):
+        correct_data.append(joggen.sensoren[i].werte)
+        incorrect_data.append(kniebeuge.sensoren[i].werte)
+    evaluate(500, amount_of_weights, correct_data, incorrect_data, perceptron)
+
 
 
 def apply_pocket():
@@ -86,37 +66,13 @@ def apply_pocket():
     train_pocket(perceptron, training_data, max_iterations, learning_rate)
     print("\nPocket: Zunächst Gewichte, dann Klassifizierung. Gehen = 1, Kniebeuge = 0")
     print(perceptron.weights)
-    cnt_true_pos = 0
-    cnt_true_neg = 0
-    cnt_false_pos = 0
-    cnt_false_neg = 0
-    unclassified = 0
-    for i in range(500):
-        jog_res = perceptron.classify(
-            [joggen.sensoren[0].werte[i], joggen.sensoren[1].werte[i], joggen.sensoren[2].werte[i]])
-        knie_res = perceptron.classify(
-            [kniebeuge.sensoren[0].werte[i], kniebeuge.sensoren[1].werte[i], kniebeuge.sensoren[2].werte[i]])
-        if math.isclose(1, jog_res, abs_tol=0.1):
-            cnt_true_pos += 1
-        elif math.isclose(0, jog_res, abs_tol=0.1):
-            cnt_false_neg += 1
-        else:
-            unclassified += 1
-        if math.isclose(0, knie_res, abs_tol=0.1):
-            cnt_true_neg += 1
-        elif math.isclose(1, knie_res, abs_tol=0.1):
-            cnt_false_pos += 1
-        else:
-            unclassified += 1
-    print(f"""
-    Anzahl Prüfungen: 1000
-    True Positive: {cnt_true_pos}
-    False Positive: {cnt_false_pos}
-    True Negative: {cnt_true_neg}
-    False Negative: {cnt_false_neg}
-    Unclassifiable: {unclassified}
-    Fehlerrate: {(cnt_false_neg+cnt_false_pos) / 10}% bzw. {(cnt_false_neg+cnt_false_pos+unclassified) / 10}%
-    """)
+    amount_of_weights = 3
+    correct_data = []
+    incorrect_data = []
+    for i in range(amount_of_weights):
+        correct_data.append(joggen.sensoren[i].werte)
+        incorrect_data.append(kniebeuge.sensoren[i].werte)
+    evaluate(500, amount_of_weights, correct_data, incorrect_data, perceptron)
 
 
 if __name__ == '__main__':

aufgaben/p6/evaluate.py

+import math
+
+from algorithm.pla.perceptron import Perceptron
+
+
+def evaluate(iterations: int, amount_of_values: int, correct_data: list, incorrect_data: list, perceptron: Perceptron):
+    cnt_true_pos = 0
+    cnt_true_neg = 0
+    cnt_false_pos = 0
+    cnt_false_neg = 0
+    unclassified = 0
+    for i in range(500):
+        corr = []
+        incorr = []
+        for j in range(amount_of_values):
+            corr.append(correct_data[j][i])
+            incorr.append(incorrect_data[j][i])
+        jog_res = perceptron.classify(corr)
+        knie_res = perceptron.classify(incorr)
+        if math.isclose(1, jog_res, abs_tol=0.1):
+            cnt_true_pos += 1
+        elif math.isclose(0, jog_res, abs_tol=0.1):
+            cnt_false_neg += 1
+        else:
+            unclassified += 1
+        if math.isclose(0, knie_res, abs_tol=0.1):
+            cnt_true_neg += 1
+        elif math.isclose(1, knie_res, abs_tol=0.1):
+            cnt_false_pos += 1
+        else:
+            unclassified += 1
+    print(f"""
+    Anzahl Prüfungen: {iterations*2}
+    True Positive: {cnt_true_pos}
+    False Positive: {cnt_false_pos}
+    True Negative: {cnt_true_neg}
+    False Negative: {cnt_false_neg}
+    Unclassifiable: {unclassified}
+    Fehlerrate: {(cnt_false_neg+cnt_false_pos) / 10}% bzw. {(cnt_false_neg+cnt_false_pos+unclassified) / 10}%
+    """)
\ No newline at end of file